Transverse spin transfer in proton–proton collisions
The NLO hard parts of the transverse spin transfer in proton–proton collisions, $a^\uparrow b\to c^\uparrow+X$: a transversely polarized parton of the polarized proton scatters and the transverse spin of the observed parton is measured, as for a transversely polarized hyperon, the spin asymmetry of a single hadron or a dihadron inside a jet, or one- and two-point energy correlators in the jet. The results are related to the double-transverse cross sections by a generalized crossing symmetry found in this calculation. new
What is computed
The incoming parton $a$ carries the density $\tfrac12\not k_a\gamma_5\not s_a$, the observed parton $c$ is projected with $\not p\,\gamma_5\not s_c$ ($\sigma(s_c)-\sigma(-s_c)$), and the cross section is $\sigma = XX\,\cos\varphi_a\cos\varphi_c + NN\,\sin\varphi_a\sin\varphi_c$, with both spins in the scattering plane ($XX$) or along its normal ($NN$). Both components are written in the distribution form below.
For a measured parton of momentum $p$ the partonic cross section $\sigma = s\,E\,d\sigma/d^3p$ is written in $v = 1 + t/s$ and $w = -u/(s+t)$, with $s=(k_a+k_b)^2$, $t=(k_a-p)^2$, $u=(k_b-p)^2$ for incoming partons of momenta $k_a$, $k_b$ and $0 < v,w < 1$:
$$\sigma = \sigma_{\rm Born}\,\delta(1-w) + P(v,w)\,c_{\rm reg}(v,w) + P(v,1)\big[c_{0}(v)\,\big[\tfrac{1}{1-w}\big]_+ + c_{1}(v)\,\big[\tfrac{\ln(1-w)}{1-w}\big]_+ + c_{\delta}(v)\,\delta(1-w)\big],$$Here $\sigma_{\rm Born}$ is the lowest-order term, and $c_{\rm reg}$, $c_0$, $c_1$, $c_\delta$ are the coefficients $R$, $B$, $C$, $A$ of the paper, with $P$ its normalization $\mathcal N$. The scales enter through $L_R=\ln(\mu_R^2/s)$, $L_F=\ln(\mu_F^2/s)$ for the incoming partons and $L_{FF}=\ln(\mu_{FF}^2/s)$ for the fragmenting one ($\mu_{FF}$ is the $\mu_F'$ of the paper).
Here $P = \alpha_s^3/(\pi s v w)$; their finite coefficients turn out equal in every class (they differ in $D$ dimensions by evanescent terms), so the transfer depends on $\cos(\varphi_a-\varphi_c)$ only. Seven classes contribute; class 10 starts at order $\alpha_s^3$. Classes 1 ($q\,q'\to q$) and 3 ($q\,\bar q'\to q$) have the same Born term but different NLO coefficients.
The calculation
As for the double-transverse process, with the transversity kernel $P^T_{qq}$ on the polarized incoming parton and on the fragmenting quark or antiquark, the unpolarized kernels on the other incoming parton, and a $D$-dimensional Born table of all seven transfer channels. The squares are computed with FeynCalc, FORM and Tracer.
A generalized crossing symmetry
This is a new result, not previously available. In the course of the calculation we found that it is nevertheless not independent of the double-transverse cross sections: the two are related by a generalized crossing symmetry. As in ordinary crossing, the transversely polarized incoming parton $b$ of $a^\uparrow b^\uparrow\to c+X$ is moved to the final state and the observed parton $c$ to the initial state. The transverse polarization is carried along: the transversity of the incoming parton $b$ becomes the measured transverse spin of the outgoing $\bar b$, with a definite transformation of the spin components and of the color and polarization averages:
$$a^\uparrow b^\uparrow\to c+X\quad\longleftrightarrow\quad a^\uparrow\bar c\to\bar b^\uparrow+X,\qquad s\leftrightarrow t,\quad u\to u,\quad v\to-\frac{v}{1-v},\quad w\to w.$$The symmetry is exact: it holds in $D$ dimensions and before any integration, for the squared matrix elements of the Born, virtual (one-loop) and real-emission contributions alike. With $\mathcal C$ the crossing of these unintegrated expressions and $T^{\rm in}$ the components of the double-transverse squares, the spin components transform as
$$T_{XX} = -\varsigma\, r\,\mathcal C\big[T_{XX}^{\rm in}\big],\qquad T_{NN} = +\varsigma\, r\,\mathcal C\big[T_{NN}^{\rm in}\big],$$with $(\varsigma,r)=(1,1)$ when the unpolarized incoming parton $\bar c$ of the transfer process is a quark or antiquark and $(-1,[(D-2)C_F]^{-1})$ when it is a gluon (its color and polarization average), kept exact in $D$ until the Laurent expansion.
Here $\varsigma$ is the fermion crossing sign. The spin factor of a quark ($\not k$, $\not p$, or $\not k\gamma_5\not s$) is linear in its momentum, which crossing reverses, so each crossed quark brings a factor $-1$; the polarization sum of a gluon is even in its momentum and brings none. The crossed parton $b$ is always a quark, since a gluon has no transversity, so $\varsigma=(-1)^2=1$ when $\bar c$ is a quark or antiquark and $\varsigma=-1$ when it is a gluon. The factor $r$ is the change of the color and polarization averages. The opposite signs of the two components come from the spin axes: the crossed in-plane axis reverses the azimuth, $\varphi\to\pi-\varphi$, which gives $-1$ for $XX$, while the normal enters both spin vectors and its reflection cancels in $NN$. Consequently the double-transverse projection $(T^{\rm in}_{XX}-T^{\rm in}_{NN})/2$ becomes the transfer projection $(T_{XX}+T_{NN})/2$. The symmetry also connects a property of each calculation to the other: the symmetric part $(T^{\rm in}_{XX}+T^{\rm in}_{NN})/2$ of the double-transverse squares is evanescent and leaves no finite remainder, which after crossing is the statement that the transfer components $XX$ and $NN$ have equal finite coefficients, so that the transfer depends on $\cos(\varphi_a-\varphi_c)$ only.
Step 5 of the transfer workflow (5-Crossing.wl) uses the symmetry as a second, independent derivation of
all seven classes from the double-transverse squares:
- the crossed Born terms and Born table equal those of the direct calculation identically;
- the crossed one-loop correction equals the direct one identically after reduction to master integrals;
- the crossed real-emission squares equal the direct ones at three random exact rational points (they are too large for a symbolic difference);
- integrated in the transfer region and combined with the counterterms of the transfer process, the crossed squares give pole-free finite coefficients equal to those of the direct calculation in exact arithmetic at three rational points of $s$, $v$, $w$ and $N_f$, with the three scales symbolic.
Direct and crossed squares pass through the same scalar reduction and finite-subtraction implementation. Since the double-transverse classes agree with the code of Mukherjee, Stratmann and Vogelsang, the symmetry ties the new transfer results to an independent published calculation.
Checks
| Class | Process | Born | Crossed from TT class | FeynCalc = FORM = Tracer | Crossing |
|---|---|---|---|---|---|
| 1 | $q\,q'\to q$ | yes | 5 | agree (2 random exact points) | Born and virtual identical, real squares at 3 exact points; NLO coefficients at 3 exact points |
| 3 | $q\,\bar q'\to q$ | yes | 5 | agree (2 random exact points) | Born and virtual identical, real squares at 3 exact points; NLO coefficients at 3 exact points |
| 6 | $q\,q\to q$ | yes | 11 | agree (2 random exact points) | Born and virtual identical, real squares at 3 exact points; NLO coefficients at 3 exact points |
| 10 | $q\,g\to \bar q$ | no | 7 | agree (2 random exact points) | Born and virtual identical, real squares at 3 exact points; NLO coefficients at 3 exact points |
| 11 | $q\,\bar q\to q$ | yes | 11 | agree (2 random exact points) | Born and virtual identical, real squares at 3 exact points; NLO coefficients at 3 exact points |
| 13 | $q\,g\to q$ | yes | 12 | agree (2 random exact points) | Born and virtual identical, real squares at 3 exact points; NLO coefficients at 3 exact points |
| 17 | $\bar q\,q\to q$ | yes | 6 | agree (2 random exact points) | Born and virtual identical, real squares at 3 exact points; NLO coefficients at 3 exact points |
All seven classes are reduced and assembled separately in each of the three algebra routes. The squared matrix elements are compared at two exact-rational points, and the finite coefficients at six sampled points. Inside the notebooks, poles cancel, the scale derivatives are those of the evolution equations, the mixed components $XN$, $NX$ vanish, the real emission is gauge invariant, and the Born table equals the lowest-order Collins hard factors of Yuan (Phys. Rev. Lett. 100 (2008) 032003, Eq. (6)) for the channels given there.
Files
Notebooks: Processes/HadronicTransfer (the crossing is
5-Crossing.wl); conventions: docs/TRANSVERSE-PP.md.
Coefficients: Coefficients/ProtonProtonTransfer.wl
(classes 1, 3, 6, 10, 11, 13, 17).
python3 Coefficients/evaluate.py transfer 1 XX 0.3 0.2 --digits 30